Kareem A. Zaghloul, Michael B. Manookin, Bart G. Borghuis, Kwabena... and Jonathan B. Demb

Kareem A. Zaghloul, Michael B. Manookin, Bart G. Borghuis, Kwabena... and Jonathan B. Demb
Kareem A. Zaghloul, Michael B. Manookin, Bart G. Borghuis, Kwabena Boahen
and Jonathan B. Demb
J Neurophysiol 97:4327-4340, 2007. First published Apr 25, 2007; doi:10.1152/jn.01091.2006
You might find this additional information useful...
This article cites 84 articles, 49 of which you can access free at:
http://jn.physiology.org/cgi/content/full/97/6/4327#BIBL
Updated information and services including high-resolution figures, can be found at:
http://jn.physiology.org/cgi/content/full/97/6/4327
Additional material and information about Journal of Neurophysiology can be found at:
http://www.the-aps.org/publications/jn
This information is current as of June 12, 2007 .
Downloaded from jn.physiology.org on June 12, 2007
Journal of Neurophysiology publishes original articles on the function of the nervous system. It is published 12 times a year
(monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 2005 by the
American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.
J Neurophysiol 97: 4327– 4340, 2007.
First published April 25, 2007; doi:10.1152/jn.01091.2006.
Functional Circuitry for Peripheral Suppression in Mammalian Y-Type
Retinal Ganglion Cells
Kareem A. Zaghloul,1 Michael B. Manookin,3 Bart G. Borghuis,1 Kwabena Boahen,2 and
Jonathan B. Demb1,3,4,5
1
Departments of Neuroscience and 2Bioengineering, University of Pennsylvania, Philadelphia, Pennsylvania; and 3Neuroscience Program
and 4Departments of Ophthalmology and Visual Sciences and 5Molecular, Cellular and Developmental Biology, University of Michigan,
Ann Arbor, Michigan
Submitted 11 October 2006; accepted in final form 22 April 2007
The retina represents a model system for understanding how
neural circuitry generates the receptive field properties of
sensory neurons. Ganglion cells are the output neuron of the
retina and have been well characterized at the level of extracellular recording. A receptive field typically features a center
region, which can be excited by an effective stimulus (EnrothCugell and Robson 1966; Kuffler 1953). For an ON-center cell,
the effective stimulus is a brightening over the center, whereas
for an OFF-center cell, the effective stimulus is a dimming over
the center. Under cone-driven conditions, the receptive field
center primarily corresponds to a feed-forward pathway from
cones to bipolar cells to the ganglion cell (Fig. 1) (Masland
2001; Sterling and Demb 2004; Wässle 2004). The ganglion
cell center shows an approximately Gaussian spatial profile,
which arises from the dome-like distribution of bipolar synapses onto the ganglion cell dendritic tree (Kier et al. 1995).
The receptive field center response is suppressed by stimulation of the surround (Kuffler 1953) (Fig. 1). In the classical
description, the surround represents a broad region extending
across and beyond the center (Enroth-Cugell and Robson 1966;
Rodieck 1965). This classical surround exhibits poor spatial
resolution and therefore senses low spatial frequencies, including broad-field stimuli and low spatial-frequency gratings. In
addition to the classical surround, the ganglion cell center can
be suppressed by high spatial frequency contrast patterns in the
peripheral receptive field. This peripheral suppression can be
distinguished from the classical surround based on spatial
resolution. For example, the spiking response to a central spot
can be suppressed by peripheral contrast, such as high spatial
frequency drifting gratings (⬃1 cycle/° or ⬃4 –5 cycles/mm on
the retina), that would not stimulate the classical surround
(Caldwell and Daw 1978a,b; Enroth-Cugell and Jakiela 1980;
Lankheet et al. 1992; Shapley and Victor 1979; Solomon et al.
2006). Peripheral suppression is not equally strong in all cell
types but is particularly prominent in the brisk-transient cell
types, which include the parasol/magnocellular pathway cells
in primates and ␣/Y-type cells in other mammals (Caldwell
and Daw 1978a,b; Enroth-Cugell and Jakiela 1980; Shapley
and Victor 1979; Solomon et al. 2006).
Surround mechanisms are created at both synaptic layers in
the retina (Fig. 1). At the outer plexiform layer, horizontal cells
inhibit cones and bipolar cells (Fig. 1) (Duebel et al. 2006;
Lankheet et al. 1992; Mangel 1991; McMahon et al. 2004).
Horizontal cells are electrically coupled to one another in a
syncytium (Baldridge et al. 1998). Thus the horizontal cell
network has requisite properties to convey the classical surround mechanism, sensitive to low spatial frequency stimuli.
At the inner plexiform layer, amacrine cells inhibit bipolar cell
synaptic terminals and ganglion cell dendrites (Fig. 1). Amacrine cells are driven by bipolar cells, which themselves have
relatively narrow receptive fields that would yield sensitivity to
high spatial frequencies (Dacey et al. 2000). Furthermore, a
nonlinearity at the bipolar cell synapse creates a subunit structure in postsynaptic cells and extends sensitivity to high spatial
frequencies (Demb et al. 2001a; Hochstein and Shapley 1976).
Thus an amacrine cell mechanism for surround inhibition
Address for reprint requests and other correspondence: J. Demb, Univ. of
Michigan, Kellogg Eye Center, 1000 Wall St., Ann Arbor, MI 48105 (E-mail:
[email protected]).
The costs of publication of this article were defrayed in part by the payment
of page charges. The article must therefore be hereby marked “advertisement”
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
INTRODUCTION
www.jn.org
0022-3077/07 $8.00 Copyright © 2007 The American Physiological Society
4327
Downloaded from jn.physiology.org on June 12, 2007
Zaghloul KA, Manookin MB, Borghuis BG, Boahen K, Demb JB.
Functional circuitry for peripheral suppression in mammalian Y-type
retinal ganglion cells. J Neurophysiol 97: 4327– 4340, 2007. First
published April 25, 2007; doi:10.1152/jn.01091.2006. A retinal ganglion cell receptive field is made up of an excitatory center and an
inhibitory surround. The surround has two components: one driven by
horizontal cells at the first synaptic layer and one driven by amacrine
cells at the second synaptic layer. Here we characterized how amacrine cells inhibit the center response of ON- and OFF-center Y-type
ganglion cells in the in vitro guinea pig retina. A high spatial
frequency grating (4 –5 cyc/mm), beyond the spatial resolution of
horizontal cells, drifted in the ganglion cell receptive field periphery
to stimulate amacrine cells. The peripheral grating suppressed the
ganglion cell spiking response to a central spot. Suppression of
spiking was strongest and observed most consistently in OFF cells. In
intracellular recordings, the grating suppressed the subthreshold membrane potential in two ways: a reduced slope (gain) of the stimulusresponse curve by ⬃20 –30% and, in OFF cells, a tonic ⬃1-mV
hyperpolarization. In voltage clamp, the grating increased an inhibitory conductance in all cells and simultaneously decreased an excitatory conductance in OFF cells. To determine whether center response
inhibition was presynaptic or postsynaptic (shunting), we measured
center response gain under voltage-clamp and current-clamp conditions. Under both conditions, the peripheral grating reduced center
response gain similarly. This result suggests that reduced gain in the
ganglion cell subthreshold center response reflects inhibition of presynaptic bipolar terminals. Thus amacrine cells suppressed ganglion
cell center response gain primarily by inhibiting bipolar cell glutamate
release.
4328
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
flattened eye cup, which included the neural retina attached to the
pigment epithelium, choroid, and sclera, was mounted on a piece of
circular filter paper and mounted flat in a chamber on a microscope
stage. The filter paper was held in place by a teflon ring that fit tightly
in the chamber walls. The tissue was superfused (⬃4 – 6 ml/min) with
oxygenated (95% O2-5% CO2) Ames medium (Sigma, St. Louis, MO)
at 32–36°C (Demb et al. 1999). In some experiments, glucose was
added to the medium (0.8 –3.6 g/l); this increased osmolarity by
⬃2– 6% and did not have noticeable effects on the recordings.
Electrophysiology
would be sensitive to both low and high spatial frequencies
(Cook and McReynolds 1998; Cook et al. 1998; Demb et al.
1999; Flores-Herr et al. 2001; Taylor 1999).
Here, we asked how amacrine cell inhibition acts to suppress
the center response of a ganglion cell. To this end, we presented high spatial frequencies in the ganglion cell’s receptive
field periphery to selectively stimulate amacrine pathways.
Direct measurements showed that such frequencies only minimally stimulate horizontal cells. Therefore the suppressive
effects measured in the ganglion cell must be driven by
amacrine circuitry. An amacrine cell’s inhibitory signal can
reach the ganglion cell either directly, through a synapse on the
dendrite, or indirectly, through a synapse onto a presynaptic
bipolar terminal (Cook and McReynolds 1998; Cook et al.
1998; Flores-Herr et al. 2001; Shields and Lukasiewicz 2003;
Taylor 1999). Here, we tested the relative contributions of
these two synaptic sites and determined their individual roles in
the suppression of the excitatory center response of the ganglion cell.
METHODS
Tissue preparation
For each experimental session, a retinal whole mount (flattened eye
cup) was prepared from a Hartley guinea pig (200 – 800 g). Procedures
were in accordance with University of Pennsylvania, University of
Michigan, and National Institutes of Health guidelines. In some cases,
an animal was anesthetized with ketamine (100 mg/kg), xylazine (20
mg/kg), and pentobarbital (150 mg/kg), and both eyes were removed,
after which the animal was killed by anesthetic overdose. In other
cases, an animal was anesthetized with ketamine (40 mg/kg) and
xylazine (4 mg/kg) and decapitated, and both eyes were removed.
Each eye was hemisected, and the vitreous was peeled off in one
piece. Five slits were cut, so that the eye cup would lay flat. The
J Neurophysiol • VOL
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
FIG. 1. Circuitry for receptive field center and surround mechanisms.
Excitatory receptive field (RF) center response of the ganglion cell (g) arises
from cone photoreceptors (c) releasing glutamate onto bipolar cells (b), which
in turn release glutamate onto the ganglion cell (g). A split into ON and OFF
pathways arises at the cone synapse, where postsynaptic bipolar cells express
either ionotropic glutamate receptors (OFF pathway), which preserve the sign of
cone response, or metabotropic receptors (ON pathway), which invert the sign
of cone response (Sterling and Demb 2004). Surround arises partly from
cone-driven horizontal cell (h) feedback. Horizontal cells excite neighboring
cells through gap junctions and ultimately inhibit cones and bipolar cells that
make up the center pathway. Surround also arises from bipolar cell-driven
amacrine cell (a) feedback. Amacrine cell-mediated inhibition can act directly
onto the ganglion cell and onto the presynaptic bipolar terminal. Shown is the
circuit for an OFF-center ganglion cell; the ON-center cell would have a similar
circuit except that cone synapses are hyperpolarizing.
For intracellular recordings of ganglion cells, Acridine orange
(0.001%; Molecular Probes, Eugene, OR) was added to the superfusate, allowing ganglion cell somas to be identified by fluorescence
during brief exposure to UV light. A soma in the visual streak was
penetrated with a glass electrode (tip resistance, 80 –200 M⍀) filled
with 1% pyranine (Molecular Probes) in 2 M potassium acetate.
Voltage was recorded with an intracellular amplifier (NeuroData,
IR-283, NeuroData Instruments, Delaware Water Gap, PA) and digitized at 5 kHz using AxoScope software (Axon Instruments, Foster
City, CA). For horizontal cells, recordings were obtained with glass
electrodes (impedance, 80 –200 M⍀) back-filled with Alexa 488 or
568 and neurobiotin (5%) in 1.5 M potassium acetate. Membrane
potential was amplified (NeuroData, IR-283), sampled at 2 kHz, and
digitized by a computer for on-line analysis and storage (Apple
Macintosh; 10-bit AD board; custom built software). Resting potential
was recorded after cell penetration and continuously monitored during
the recording. After the recording, horizontal cells were injected with
the dye and fixed for confocal microscopy.
For extracellular (loose-patch) and whole cell recordings from
ganglion cells, we identified large somas visualized with a cooled
CCD camera (Retiga 1300C, QCapture software, Qimaging, Burnaby,
British Columbia, Canada) (Hu et al. 2000). Positive pressure was
applied through a patch electrode (3– 6 M⍀) and filled with Ames
medium, and this electrode was used to “burrow” through the inner
limiting membrane and clean the area surrounding the targeted cell
(Roska and Werblin 2001). The same electrode was used to form a
loose seal onto the targeted cell (⬃30 –100 M⍀) and record spikes
under voltage clamp (Vhold ⫽ 0 mV). This electrode was withdrawn,
and a second electrode (3– 6 M⍀) filled with intracellular solution was
used to obtain a ⬎1-G⍀ seal. The patch was ruptured, and recordings
were made in the whole cell configuration. Intracellular solution for
current-clamp recordings contained (in mM) 140 K methylsulfate, 8
NaCl, 10 HEPES, 0.1 EGTA, 2 ATP-Mg, and 0.3 GTP-Na2, adjusted
to pH 7.3. For voltage-clamp recordings, QX-314-Br was added (5
mM) to block sodium channels and improve space clamp, and NaCl
was reduced to 3 mM. The calculated reversal for inhibitory responses
(EGABA/glycine) would be approximately ⫺73 mV for the first solution.
For the second solution, containing Br⫺, EGABA/glycine should be more
depolarized given that GABA/glycine channels are more permeable to
Br⫺ than Cl⫺ by ⬃20% (Bormann et al. 1987; Robertson 1989). In
calculations below, we assumed that with Br⫺ in the pipette, EGABA/
glycine was approximately ⫺68 mV. Junction potential (⫺9 mV) was
corrected in all cases.
For voltage-clamp recordings, the holding potential was compensated for the voltage drop across the electrode tip, based on the
following equation: Vh_corr ⫽ Vh – [Ileak Rs (1 – Rs_correct)], where Vh
is the apparent holding potential before the stimulus (in mV), Ileak is
the leak current (in nA), Rs is the series resistance (typically 12–30
M⍀), and Rs_correct is the series resistance compensation (typically
between 0.25 and 0.50). Holding potentials were typically restricted to
a range negative to ⫺30 mV; positive to ⫺30, the cells showed a large
outward current, which resembled a delayed-rectifier potassium current. Signals were recorded with a MultiClamp 700A amplifier and
digitized at 10 kHz using pClamp 9.0 software (Axon Instruments).
RETINAL CIRCUITRY FOR PERIPHERAL SUPPRESSION
Programs were written in Matlab (Mathworks, Natick, MA) to
analyze responses (downsampled to 1 kHz) separately in the spike rate
and subthreshold membrane potential, as described previously (Demb
et al. 2001a; Zaghloul et al. 2003, 2005). To exclude the spikes from
the voltage response, we down-sampled the raw data (recorded at 5 or
10 kHz) by taking the median value of every 5 or 10 sample points;
this had the effect of removing spikes but preserving the subthreshold
waveform. In some cells, we instead used a linear interpolation
method, where we clipped out each spike with a line from 1 ms before
to 1–2 ms after each spike (Zaghloul et al. 2003).
Data are reported as mean ⫾ SE. Statistical significance was
assessed using a one-tailed t-test.
Horizontal cell morphology
Visual stimulus
The stimulus was displayed on a miniature monochrome computer
monitor (Lucivid MR1-103, Microbrightfield, Colchester, VT) projected through a microscope port and through a ⫻2.5 or ⫻4 objective
focused on the photoreceptors. The monitor had a vertical refresh of
60 Hz and a spatial resolution of 640 ⫻ 480. The stimulus was
confined to the central 480 ⫻ 480 pixels, which, when projected onto
the retina, subtended 3 ⫻ 3 or 3.7 ⫻ 3.7 mm. The mean luminance
was ⬃103-104 isomerizations per middle-wavelength sensitive cone
or rod per second, which is within the mesopic rang (Yin et al. 2006);
in two cases, a lower mean luminance was used (⬃102 isomerizations
per cone or rod per second). The relationship between gun voltage and
monitor intensity was linearized in software with a lookup table.
Stimuli were programmed in Matlab as described previously (Brainard 1997; Demb et al. 1999; Pelli 1997).
All ganglion cell experiments used a dynamic modulation of a lowcontrast spot (diameter, 0.5 mm) centered on the cell body and therefore
approximately centered on both the dendritic tree and the ⬃0.5- to
0.7-mm-diam receptive field center (Demb et al. 2001a,b; Dhingra et al.
2003). The spot will necessarily stimulate both the classical center and the
surround, because these overlap in space (Enroth-Cugell and Robson
1966). However, the purpose of the spot was to stimulate the bipolar cells
that synapse onto the ganglion cell dendritic tree, and for this purpose, the
spot, even if centering was off by ⬃0.1 mm, would be adequate to
stimulate most of these bipolar cells.
The spot intensity was updated at 60 Hz, with values drawn
randomly from a Gaussian distribution. This stimulus approximates
“white noise” (Marmarelis and Marmarelis 1978; Sakai and Naka
1987; Fig. 2). The stimulus distribution is presented in contrast units,
where the intensity has been normalized by subtracting the mean
luminance and dividing by the mean luminance. Thus the stimulus had
a mean of 0 and a range of ⫺1 to ⫹1. Stimulus contrast is defined by
the Gaussian SD, which was always 0.10. The stimulus lasted 240 s
and included twelve 20-s periods: 10 s of the spot (spot alone)
alternating with 10 s of the spot plus a drifting grating (spot ⫹
grating). The grating was presented in the periphery, excluded from a
1-mm patch centered on the cell body. The grating had a square-wave
profile with a spatial frequency of 4.3 or 5.0 cycles/mm (bar width ⫽
116 or 100 ␮m), a temporal frequency of 2 Hz, and a contrast of 1.0.
The analysis was performed on data collected during the last 8 s of
each 10-s half-period. Thus for each contrast, there was 8 ⫻ 12 ⫽ 96 s
of data. This relatively short stimulus increased the probability of
highly stable intracellular recordings and provided data with sufficient
signal-to-noise for the analysis. Because of the alternating contrast
half-periods, any instabilities during the recording should be distributed equally between the two half-periods. Horizontal cell recordings
used drifting sine-wave gratings (2 Hz, 0.7 contrast, 0.05– 6 cycles/
mm) presented over a field of 2.4 ⫻ 3.2 mm.
Analysis: linear-nonlinear model
We analyzed both subthreshold and spiking responses using a
linear-nonlinear (LN) cascade analysis (Carandini et al. 2005; Chichilnisky 2001). In this analysis, a linear filter represents the impulse
response function of a cell, or the theoretical response to a brief light
flash (Fig. 2). This filter, plotted in reverse, represents the cells
weighting function. The linear prediction of the response (i.e., the
linear model) is calculated at a given point in time by multiplying the
stimulus by the weighting function and summing the result (i.e.,
convolution; Carandini et al. 2005). To compute the L filter, f(t), we
cross-correlated the stimulus and the response (Chichilnisky 2001;
Lee and Schetzen 1965; Sakai et al. 1995; Wiener 1958). To generate
the L model of response (rL), we convolved (*) the stimulus [s(t)] and
the L filter
r L(t) ⫽ f(t)*s(t)
Finally, the L model is passed through a nonlinear input–output
function to generate a LN model of the response (Fig. 2)
FIG. 2. Linear-nonlinear (LN) model of spiking and subthreshold membrane potential responses. Stimulus (in contrast units) is convolved with a linear filter
to generate linear model of response. Linear model (in input units) is passed through an instantaneous (static) nonlinearity to generate the LN model of the
response (in spikes/s or mV). Model (green lines), built from 1 data set, can be compared with average response to a brief test stimulus (black or red lines) from
a separate data set (averaged over 24 repeats to reduce noise). Model corresponds closely to data. Shown are models for spiking (top row) and subthreshold
membrane potential responses (bottom row) to spot alone and spot ⫹ grating conditions. The r2 between model and data was 0.88 (spikes, spot alone); 0.67
(spikes, spot ⫹ grating); 0.95 (membrane potential, spot alone); and 0.91 (membrane potential, spot ⫹ grating).
J Neurophysiol • VOL
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
After recording horizontal cells, the retina was fixed in 4% paraformaldehyde for 20 min. Cells were visualized either based on the
Alexa dye staining or by reacting for Neurobiotin by the following
procedure: incubate with blocking buffer (10% normal goat serum,
5% Triton-X in 5% sodium phosphate buffer, 1 h); react with
streptavidin-fluorescein (8 mg/ml) or streptavidin-Cy5 (40 mg/ml) for
3 h; wash with 5% sodium phosphate buffer (3 ⫻ 10 min); mount on
a slide using Vectashield (Vector Laboratories, Burlingame, CA).
Cells were imaged using a confocal microscope (Leica, Nussloch,
Germany; ⫻40 objective; NA, 1.25).
4329
4330
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
r LN(t) ⫽ N关rL共t兲兴
The N function accounts for rectification (e.g., the spike threshold)
and saturation in the response (Baccus and Meister 2002; Chichilnisky
2001; Kim and Rieke 2001; Sakai et al. 1995; Victor 1987; Zaghloul
et al. 2003).
To generate the N function, we plotted the L model versus the
actual response (analyzed after down-sampling to 1 kHz) and binned
the data (⬃1000 samples/bin). These binned points represent the
average response output (in spikes/s, mV, or pA) at each level of the
L model value (in arbitrary input units). For the simultaneous fitting
procedure described below, we required a descriptive function fitted
through the points of the N function. A Gaussian cumulative distribution function (cdf) provides a good fit to the spike N function
(Chander and Chichilnisky 2001; Zaghloul et al. 2003, 2005)
f(x) ⫽ ␣C(␤x ⫹ ␥)
f(x) ⫽ ␣C(␤x ⫹ ␥) ⫹ ␦
where the additional parameter ␦ allows for a vertical offset (because
the membrane N function goes negative). As we found previously, the
Gaussian cdf provided a good fit for the membrane potential N
function for ON cells but not for OFF cells (Zaghloul et al. 2003). Thus
for OFF cells, we fit the positive and negative sides of the membrane
N function separately (Zaghloul et al. 2003, 2005).
Analysis: simultaneous fit of subthreshold responses to the
spot alone and spot ⫹ grating conditions
For membrane potential or current responses, we fit both low and
high contrast N functions with the Gaussian cdf plus offset. The low
and high contrast functions were fit with the same ␣ and ␥ parameters
but unique ␤ and ␦ parameters. Thus the fitted N functions were
similar to those for the spike N functions, except that membrane
potential N functions were allowed to have unique offsets (i.e., unique
y-intercepts because of the unique ␦). The unique offsets were necessary, because in some cases, the y-intercept of the membrane N
function differed between contrasts by several millivolts or tens of
picoamperes (Baccus and Meister 2002). Furthermore, the gain
change between conditions (change in ␤) was independent of the
change in the y-intercept. For ON cells, we performed the simultaneous
fit with the cdf to the full nonlinear function, whereas for OFF cells, we
fit just the excitatory (depolarizing or inward current) side of the
nonlinear function (Zaghloul et al. 2003, 2005). For the subthreshold
response, the above gain change model showed good fits, without
systematic errors of any sort, and so we did not consider alternative
models as we did for the spike response.
Analysis: simultaneous fit of spike responses to the spot
alone and spot ⫹ grating conditions
Evaluation of the LN model
There is a free scale factor in the LN model: the y-axis of the filter
and the x-axis of the nonlinear function can be scaled by the same
amount without changing the output of the model (Chander and
Chichilnisky 2001; Kim and Rieke 2001; Zaghloul et al. 2005).
Therefore to compare between the two conditions, we initially multiplied each filter by a factor, so that the variance of the two linear
models was equal (Baccus and Meister 2002). We scaled the nonlinearities by the same factors (i.e., a stretch along the x-axis) so that
each LN model output remained unchanged. After normalizing in this
way, the two filters showed only slight differences in kinetics, but
there was a large difference between the N functions: the peripheral
grating typically reduced the spike rate, which resulted in an apparent
rightward shift of the N function.
We considered three models to explain the effect of the peripheral
grating on the N stage of the model. First, we considered a gain
change model in which the two curves were fit with the same
Gaussian cdf, except for a scaling along the x-axis. Specifically, both
curves were fit with the same ␣ and ␥ parameters, but unique ␤
parameters (i.e., total of four parameters). This model was used
previously to account for the reduced gain evoked by an increase in
the contrast of the spot itself (Beaudoin et al. 2007; Chander and
Chichilnisky 2001; Kim and Rieke 2001; Zaghloul et al. 2005).
However, based on a preliminary analysis of OFF cells, it was apparent
that this model failed in a systematic way: an underestimation of the
small amplitude responses to the spot alone and an overestimation of
the small amplitude responses in the presence of the grating. Therefore we considered a threshold change model in which the two curves
were fit with the same Gaussian cdf except for a horizontal shift along
the x-axis. Specifically, both curves were fit with the same ␣ and ␤
parameters, but unique ␥ parameters (i.e., total of four parameters).
The LN model is useful to the extent that it provides an accurate
representation of a cell’s response. We evaluated model accuracy by
building the model based on data collected with one stimulus and
testing the model’s predictive power (explained variance: r2) on the
averaged response to a second stimulus. The model-building stimulus
was a modulated spot, as described above: twelve 20-s periods (10 s
of spot alone and 10 s of spot ⫹ grating). The model-testing stimulus
was a 2-s segment of spot modulation, repeated 10 times in a 20-s
period (10 s of spot alone and 10 s of spot ⫹ grating). The modeltesting period (MT) was interspersed six times within the model
building periods (MB): 3 MT; 2 MB; 3 MT; 2 MB; 3 MT; 2 MB; and
3 MT. In all cases, we only analyzed the last 8 s of data from each 10-s
half-period. Thus the LN model was built from 96 s of data for both
spot alone and spot ⫹ grating conditions. The response to the 2-s
model-testing segment was averaged over 24 repeats for both spot
alone and spot ⫹ grating conditions. We tested how well the LN
model predicted the averaged response to the 2-s model-testing
segment (Fig. 2). The model was assessed in this way for 14 cells (11
OFF cells and 3 ON cells) and for both spiking and subthreshold
measurements. For spiking responses, we evaluated the model in
cases where the firing rate was at least 1 spike/s for both conditions.
We evaluated model fits for the theshold change model, which
generally yielded better fits than the gain change model.
For spiking responses, the r2 between the LN model and the data
were 0.80 ⫾ 0.03 for the spot alone condition and 0.65 ⫾ 0.03 for the
spot ⫹ grating condition (n ⫽ 13). For membrane potential responses,
the r2 between the LN model and the data were 0.94 ⫾ 0.01 for the
spot alone condition and 0.90 ⫾ 0.01 for the spot ⫹ grating condition
(n ⫽ 11). Thus the LN model generally gave a better fit for the spot
alone condition than for the spot ⫹ grating condition, although this
J Neurophysiol • VOL
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
where (C) is the cumulative normal density, x is the input value, and
the parameters roughly correspond to the maximum response (␣), the
response gain (␤), and the spike threshold (␥) (Chander and Chichilnisky 2001; Chichilnisky 2001). The fit was performed using standard
routines in Matlab that minimize the mean-squared error (MSE)
between the data and the fitted line (Fig. 2). For the membrane N
points, we fit the data with the function
Finally, we considered a combined model that allowed for both a gain
change and a threshold change. Specifically, both curves were fit with
the same ␣ parameters but unique ␤ and ␥ parameters (i.e., total of 5
parameters).
The gain change and threshold change models have the same form
and equal number of parameters so we could directly compare the
MSE of these models to determine which provided a better fit to the
data. The combined model has an additional parameter and so we
expect it to yield a lower MSE. However, the combined model
allowed us to test whether the gain change or threshold change
dominated in explaining the difference between the two contrast
curves when both gain and threshold were allowed to vary.
RETINAL CIRCUITRY FOR PERIPHERAL SUPPRESSION
4331
difference was most marked for spiking responses; membrane potential responses were very well captured by the LN model in either
condition. The relatively low r2 value for spiking response to the
spot ⫹ grating condition was likely caused by the lower spike rates in
this condition.
RESULTS
Basic cellular properties
Spatial tuning of horizontal cells in the guinea pig retina
We recorded horizontal cells to assess their spatial sensitivity in the guinea pig retina. Here we report on cells that were
successfully filled with a fluorescent dye to identify the type
(A-type or B-type; Peichl and Gonzalez-Soriano 1993). At
mean luminance, horizontal cells rested at ⫺47 ⫾ 5 mV (n ⫽
30). Responses were recorded to sine-wave gratings at a range
of spatial frequencies; the response was quantified as the
amplitude of the best fitting sine-wave at the 2-Hz drift frequency (F1 amplitude). Maximal response amplitudes were
3.8 ⫾ 0.5 (A-type; n ⫽ 24) and 3.2 ⫾ 1.1 mV (B-type; n ⫽ 6).
Both types of horizontal cell showed low-pass spatial sensitivity with half-maximal sensitivities near 0.4 – 0.5 cycles/mm
J Neurophysiol • VOL
FIG. 3. Spatial bandwidth of horizontal cells. A: A-type horizontal cell in
the visual streak of guinea pig retina. Fluorescence image was converted to
grayscale, and contrast was inverted. B: intracellular responses of the cell in A
to drifting sine-wave gratings at various spatial frequencies (0.7 contrast). Each
trace shows average of 2 repetitions. Resting potential was ⫺47 mV. C: Cell’s
spatial transfer function is low-pass with a half-maximal response at ⬃0.2
cyc/mm (same cell as in A and B). Responses are F1 amplitudes at each spatial
frequency. Smooth gray line shows a sigmoidal fit (Matlab; least-squares
method). Dashed line shows F1 amplitude to a blank screen (spontaneous). D:
average spatial transfer functions for a population of A-type and B-type
horizontal cells. Responses have been normalized by subtracting spontaneous
response and dividing by maximal response. At 5 cyc/mm, responses are
attenuated to ⬍10% of the peak response.
(Fig. 3D). At 5 cycles/mm, the response declined to ⬃4 (HA
cells, n ⫽ 24) or ⬃7% (HB cells, n ⫽ 6) of the peak response
(Fig. 3D). In the following experiments, we used spatial
frequencies of 4.3 or 5.0 cycles/mm. At these frequencies, the
bar width (⬃100 ␮m) should match the putative width of a
bipolar cell receptive field, which drive amacrine cells but is
too fine to strongly stimulate horizontal cells (Fig. 3) (Dacey et
al. 2000; Demb 1999, 2001a,b; Lankheet 1992; Passaglia et al.
2001; Werblin 1972).
Peripheral contrast suppresses the ganglion cell spiking
response to a central stimulus
We first quantified the effect of the peripheral grating on the
ganglion cell receptive field center as measured in the spike
response. To stimulate the cell, a 0.5-mm-diam spot flickered
over the center; on each frame of the monitor (16.7 ms), an
intensity value was drawn randomly from a Gaussian distribution (see METHODS and Fig. 4A). This stimulus approximates
white noise, with approximately equal stimulus energy over the
range of temporal frequencies to which the cell is most sensitive (Zaghloul et al. 2005). In alternate 10-s half-cycles, the
spot was presented either alone or in the presence of a grating
in the peripheral receptive field (Fig. 4A). During the 10-s
presentation, the grating drifted at 2 Hz and thus maintained a
constant contrast signal at each point in the periphery (EnrothCugell and Jakiela 1980).
The central spot, presented alone, caused a series of spike
bursts in the ganglion cell, and this spiking was suppressed by
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
We targeted Y-type (␣) ganglion cells, because these cells
show significant peripheral suppression in vivo and because we
could identify them routinely by their large cell bodies in the
visual streak (Caldwell and Daw 1978b; Demb et al. 1999,
2001a; Enroth-Cugell and Jakiela 1980; Peichl et al. 1987). We
presented a series of stimuli to confirm that the cell had the
characteristic properties: a “brisk-transient” center response to
a spot (0.5 mm diam), an antagonistic surround response to an
annulus (inner diameter, 0.74; outer diameter, 2.0 mm), and a
nonlinear (frequency-doubled) response to a contrast-reversing
grating (spatial frequency ⫽ 4.3 or 5.0 cycles/mm) (Cleland
and Levick 1974; Demb et al. 2001a,b; Enroth-Cugell and
Robson 1966; Hochstein and Shapley 1976). In whole cell
recordings, cells had an input resistance of 32 ⫾ 3 (SE) M⍀
(n ⫽ 17), similar to our previous measurements and similar to
Y-types cells in the cat retina (Beaudoin et al. 2007; Cohen
2001; Manookin and Demb 2006; O’Brien et al. 2002;
Zaghloul et al. 2003).
Here we report on 65 ganglion cells (48 OFF-center and 17
ON-center). As in previous studies, we had a bias toward
successful recordings from OFF-center cells (Zaghloul et al.
2003, 2005), which outnumber their ON-center counterparts by
about twofold (B. Borghuis, unpublished observations). We
recorded extracellularly from 36 cells (26 OFF-center and 10
ON-center), which, when viewing a gray screen at mean luminance, fired spontaneously at 10 ⫾ 3 spikes/s (range ⫽ 0 –57).
We recorded intracellularly from 32 cells (25 OFF-center and 7
ON-center), which fired spontaneously at 5 ⫾ 2 spikes/s
(range ⫽ 0 – 49) and rested at ⫺64.1 ⫾ 0.9 mV. Ten cells were
recorded with sharp intracellular electrodes, and 22 were recorded with whole cell patch electrodes. We made additional
whole cell recordings with QX-314 (5 mM) added to the
pipette solution to block spiking (n ⫽ 13 cells). Some cells
were recorded by both extracellular and intracellular methods
(n ⫽ 23 cells). Our main criteria for stable intracellular recordings were a low resting potential and a well-modulated light
response to the flickering spot stimulus described below.
4332
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
the addition of the peripheral grating. Similar results were
observed in spiking responses recorded by extracellular and
intracellular methods, and so results on spiking, here and
below, were combined across recording conditions. We analyzed the effect of the grating on the spike rate over the last 8 s
of each 10-s half-cycle. For OFF cells, the spike rate reduced
from 12.2 ⫾ 1.3 to 5.6 ⫾ 0.8 spikes/s, a drop of 6.7 ⫾ 0.6 (n ⫽
44 cells, 26 extracellular; P ⬍ 0.001). For ON cells, the spike
rate reduced from 32 ⫾ 4 to 30 ⫾ 4 spikes/s, a drop of 1.8 ⫾
1.9 (n ⫽ 16 cells and 10 extracellular; not significant; Fig. 4D).
We further quantified the effect of the grating by testing for a
significant reduction in spiking within individual cells. For
each cell, we compared the spike rate between the two conditions across the 12 cycles and performed a t-test (2-tailed). For
every OFF cell, the grating significantly reduced the firing rate
(P ⬍ 0.01 for 41 cells; P ⬍ 0.05 for 3 cells; Fig. 4, B and D).
For ON cells, the grating significantly reduced the rate in five
cells (P ⬍ 0.05), had no significant effect in four (P ⬎ 0.05),
and significantly increased the rate in seven (P ⬍ 0.05; Fig. 4,
B and D); the increased rate in some ON cells could result from
inadvertent stimulation of the edge of the receptive field center.
Thus the grating consistently reduced the spike rate in OFF cells
but had a more variable effect in ON cells.
Peripheral contrast causes the linear filter to become more
biphasic in OFF-center cells
To further quantify the impact of the peripheral grating on
the center response, we analyzed the spike train with an LN
analysis (see METHODS and Fig. 2). The goal of the LN analysis
J Neurophysiol • VOL
was to separate changes in the temporal sensitivity of the cell
from changes in contrast sensitivity (Carandini et al. 2005;
Chichilnisky 2001). The linear filter indicates the temporal
sensitivity of the cell. The nonlinear function shows the relationship between the filtered contrast (i.e., the linear model of
the response) and the output of the cell; the slope of this
function indicates the contrast sensitivity (see METHODS). For
the spike response, we restricted our analysis to those cells that
fired at a rate of at least 1 spike/s in both conditions and cells
that showed a significant reduction in firing rate during the
spot ⫹ grating condition (n ⫽ 40 OFF cells and 5 ON cells).
For OFF cells, the grating caused the linear filter to become
more biphasic (Fig. 5C). To quantify the change in the shape of
the filter, we compared the amplitude of the first phase of the
response (i.e., positive response for ON cells; negative response
for OFF cells), with the second phase of the response (i.e.,
negative response for ON cells; positive response for OFF cells).
We calculated a biphasic index, which was the second phase
amplitude (s2 or sg2 for the spot alone or spot ⫹ grating)
divided by the first phase amplitude (s1 or sg1). If there were no
second phase of the response, the index would be zero, whereas
if the second phase amplitude equaled the first phase amplitude, the index would be ⫺1. Most cells, with or without the
peripheral grating present, had an index between ⫺0.2 and
⫺1.0 (Fig. 5C).
For OFF cells (n ⫽ 40 cells), the biphasic index was ⫺0.55 ⫾
0.02 for the spot alone condition compared with ⫺0.73 ⫾ 0.03
for the spot ⫹ grating condition, a difference of ⫺0.17 ⫾ 0.02
(P ⬍ 0.001). For ON cells (n ⫽ 5 cells), the biphasic index was
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
FIG. 4. Peripheral contrast suppresses spiking responses to a central spot. A: ganglion cell extracellular response to 1 cycle of stimulus protocol (bottom trace).
Each cycle consisted of a 10-s period of a flickering spot (0.5 mm diam, centered on ganglion cell body) followed by a 10-s period of spot plus a drifting grating
in the periphery; spot and grating time-courses are indicated above trace. Grating was excluded from a central patch (1 mm diam; see image at top). Grating was
full contrast with 100-␮m-wide bars and a 2-Hz drift rate. Grating suppressed spiking response to the spot. B: effect of grating on firing rate in individual cells.
Plotted are firing rates during the last 8 s of each half-cycle. Grating reduced firing rate in all OFF cells and in some ON cells (points below dashed identity line).
Error bars indicate ⫾SE across 12 cycles. C: firing rate across 20-s stimulus cycle, averaged across 12 cycles (same cell as A). Error bars indicate ⫾SD across
cycles. Line above data, here and in D, indicates grating time-course (as in A). D: normalized spiking response across the 20-s stimulus cycle. Normalized
response is firing rate at each point in stimulus cycle minus average rate over entire cycle. Left: average for 44 OFF cells, all of which showed a significant decrease
in firing rate during grating. Right: average of 5 ON cells that showed a decreased spike rate (gray symbols) and 7 ON cells that showed an increased spike rate
during grating (white symbols). Error bars indicate ⫾SE across cells.
RETINAL CIRCUITRY FOR PERIPHERAL SUPPRESSION
4333
⫺0.91 ⫾ 0.07 for the spot alone condition compared with
⫺0.86 ⫾ 0.09 for the spot ⫹ grating condition, a difference of
⫺0.05 ⫾ 0.04 (not significant). Thus the peripheral grating
caused the linear filter to become more biphasic in OFF cells but
had no significant effect in ON cells. The increased biphasic
quality of the OFF cell filter, caused by the peripheral grating,
should correspond to a relatively decreased sensitivity to low
temporal frequencies.
Peripheral suppression in
increased spike threshold
J Neurophysiol • VOL
cells is best explained by an
We next compared the effect of the grating on the spiking
nonlinear function. We considered two models, with an equal
number of parameters (4), to describe the suppression of
spiking caused by the grating. In the first model, the grating
causes a gain change. The reduced gain corresponds to a
reduced slope of the nonlinear function. We also considered a
second model, in which the grating causes an increased threshold for spiking (threshold change model). This increased
threshold corresponds to a rightward shift (on a linear axis) in
the nonlinear function in the spot ⫹ grating condition (see
METHODS). For an OFF cell, the threshold change model provided
a more satisfactory fit (Fig. 5A). We also considered a combined model, with an additional parameter (5), in which the
two nonlinear functions differed by both their gain (slope) and
their threshold (horizontal position along the x-axis).
To quantify the difference between the gain change model
and the threshold change model, we fit both models for each
cell and compared the difference in MSE between each model
and the data. For OFF cells (n ⫽ 40 cells), the threshold change
model yielded lower MSE (7.2 ⫾ 0.8 spikes/s) compared with
the gain change model (11.1 ⫾ 1.4 spikes/s); the MSE was
relatively lower for the threshold change model by 27 ⫾ 4%
(P ⬍ 0.001). The combined model fit yielded a lower MSE
(6.8 ⫾ 0.7 spikes/s) compared with the models above, as
expected based on the additional parameter added in the fit.
However, the MSE for the combined model was only slightly
lower than that for the threshold change model above. Furthermore, in the combined model fit, the relative gain in the
presence of the grating was 0.98 ⫾ 0.02 (i.e., reduction of
⬃2%), and thus a gain change did not play a major role in
describing the suppressed spiking. In the combined model, the
threshold change (i.e., rightward shift) was 23.8 ⫾ 1.4 arbitrary
(linear model) units, similar to that found using the threshold
change model (23.3 ⫾ 1.5 units). Thus for OFF cells, the effect
of the grating could be described concisely as an increased
spike threshold.
For ON cells (n ⫽ 5), the MSE was not significantly lower for
the threshold change model (31 ⫾ 8 spikes/s) compared with
the gain change model (57 ⫾ 27 spikes/s); the threshold change
model was lower by 19 ⫾ 14% (not significant). For the
combined model (MSE ⫽ 28 ⫾ 7 spikes/s), the peripheral
grating caused both a significant gain reduction (0.85 ⫾ 0.03;
i.e., reduction of ⬃15%) and an increased threshold (rightward
shift, 19 ⫾ 6 units). Thus for this subset of ON cells (which
showed significantly reduced firing during the grating), the
effect of the grating could be described as a combination of an
increased spike threshold and a reduced gain.
In the preceding experiments, the grating was always presented at full contrast (1.0). In six OFF cells, we further tested
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
FIG. 5. Peripheral contrast changes response kinetics and increases threshold for spiking consistently in OFF cells. A1: linear filter for spot alone and
spot ⫹ grating conditions. Peripheral grating caused the filter to become more
biphasic. To quantify biphasic nature of the filter, we measured amplitude of
the 1st phase of the filter (negative response for OFF cells, positive for ON cells)
for both the spot alone (s1) and spot ⫹ grating condition (sg1) and compared
this with amplitude of the 2nd phase of the filter (positive response for OFF
cells, negative for ON cells) for spot alone (s2) and spot ⫹ grating condition
(sg2). Filters are from the OFF cell in Fig. 4A. A2: nonlinearity for spot alone
and spot ⫹ grating conditions. Data points show binned spike rate as a function
of linear model values. Data are fit (smooth lines) with Gaussian cumulative
distribution functions (cdfs) that are identical except for a scale factor that
stretches curves along the x-axis (gain change model) or shifts curves laterally
along x-axis (threshold change model). The threshold change model fit the data
with a lower mean squared error (MSE; 4.7 spikes/s) than the gain change
model (10.7 spikes/s). The gain change model underestimated small responses
in the spot alone condition and overestimated small responses in spot ⫹ grating
condition (gray brackets). A3: residuals (difference between data and fit) from
A2. B1: same format as A1 for an ON cell. Grating did not evoke a change in
the shape of the filter. B2: same format as A2 for an ON cell. Data were better
described by the threshold change model than the gain change model in this
cell. C: population analysis of effect of peripheral grating on response kinetics.
Grating caused OFF cells to become more biphasic (points below identity line).
ON cells were relatively unaffected. Pattern of results (here and in D) was
consistent for spikes measured with extracellular (extra) and intracellular
(intra) recordings. Analysis here and in D was restricted to those cells that fired
at a rate of at least 1 spike/s for both conditions and for which grating caused
a significant suppression of spike rate (n ⫽ 40 OFF cells, n ⫽ 5 ON cells). D:
the threshold change model best explains suppression of spiking in presence of
peripheral grating for OFF cells. Plot shows MSE for the gain change and
threshold change models. Most OFF cell points fall below the identity line,
indicating that the threshold change model yields lower MSE. ON cells were
about equally well fit by the 2 models.
OFF
4334
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
membrane potential was negligible. The membrane potential
was ⫺59.7 ⫾ 2.6 mV during the spot alone and ⫺59.4 ⫾ 2.5
mV in the presence of the grating (n ⫽ 7). ON cells that showed
a depolarization during the grating corresponded to those that
showed an increased spike rate (see above and Fig. 4B),
whereas cells that showed a hyperpolarization corresponded to
those that showed a decreased spike rate (n ⫽ 5; Fig. 6C). As
discussed below, the depolarization in some ON cells probably
arises from unintended stimulation of excitatory bipolar cell
synapses at the edge of the ganglion cell’s receptive field center
(see DISCUSSION).
Peripheral contrast reduces the response gain of
subthreshold responses
FIG. 6. Peripheral contrast hyperpolarizes the membrane potential in OFF
cells. A: OFF ganglion cell intracellular response to 1 cycle of the stimulus
protocol; shown are the 1st second, the middle 2 s, and the last second of the
20-s cycle. B: average subthreshold membrane potential across the 20-s
stimulus cycle (same cell as A). Grating caused a hyperpolarization that
partially recovered during the 10-s half-cycle. Error bars indicate ⫾SD across
cycles. C: normalized membrane potential across the 20-s stimulus cycle.
Normalized response is membrane potential at each point in stimulus cycle
minus average membrane potential over entire cycle. Left: average for 25 OFF
cells. Right: average of 2 ON cells that showed a hyperpolarization (gray
symbols) and 3 ON cells that showed a depolarization during the grating (white
symbols). Error bars indicate ⫾SE across cells.
the effect of the grating at a lower contrast (0.25). Data were
analyzed using the threshold change model. For these cells, the
low contrast grating caused a rightward shift of the nonlinear
function of 8 ⫾ 2 linear model units at low contrast and 20 ⫾
3 units at high contrast. Thus there was a significant effect of
the grating at low contrast with a significantly greater effect at
high contrast (difference of 12 ⫾ 2 units; P ⬍ 0.01). Therefore
at the lower contrast level, the effect of the grating was not
saturated.
Peripheral contrast causes tonic membrane hyperpolarization
in OFF cells
To understand the mechanism underlying the above effects
on spiking, we analyzed the effect of the peripheral grating on
the subthreshold membrane potential. In OFF cells, the grating
evoked a tonic membrane hyperpolarization (Fig. 6). The
hyperpolarization was largest at the onset of the peripheral
stimulus and slowly declined over the 10-s half-cycle. During
the period of analysis (2–10 s after grating onset or offset), the
grating hyperpolarized the membrane potential from ⫺64.6 ⫾
0.9 to ⫺65.7 ⫾ 0.8 mV, a difference of 1.2 ⫾ 0.2 mV (n ⫽ 25
cells; P ⬍ 0.001). At the offset of the peripheral stimulus, the
membrane initially depolarized strongly, and this depolarization declined during the half-cycle (Fig. 6C).
Individual ON cells showed either a tonic depolarization or
hyperpolarization, so that the average effect of the grating on
J Neurophysiol • VOL
FIG. 7. Peripheral contrast reduces gain of subthreshold responses. A1:
linear filter and static nonlinearity for membrane potential response to spot
alone and spot ⫹ grating conditions. Inset: linear filters for the 2 conditions
(normalized to have equal variance; scale bar indicates 100 ms; y-axis is in
arbitrary filter units). Points show the 2 static nonlinearities. Depolarizing
(rightward) sides of nonlinearities were fit with Gaussian cdfs that differed by
a gain change (i.e., scaling along the x-axis) and a tonic hyperpolarization (i.e.,
vertical shift, reflected by a drop in y-intercept). Grating reduced gain by 36%
and hyperpolarized membrane by 2.6 mV. Data are a whole cell recording of
the OFF cell in Fig. 6A. A2: analysis of spiking response for cell in A1.
Threshold change model was fit to data (rightward shift of 32 input units). B1:
same as A1 for an ON cell. Grating reduced gain by 23% and hyperpolarized
membrane by 1.0 mV. B2: analysis of spiking response for the cell B1.
Combined model was fit to data (reduced gain by 21%; rightward shift of 12
input units). C1: same as B1 for a 2nd ON cell. Grating reduced gain by 21%
and depolarized membrane by 0.4 mV. C2: analysis of spiking response for
cell in C1. Combined model was fit to data (reduced gain by 27%, leftward
shift of 1 input unit).
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
In addition to tonic hyperpolarization of the membrane
potential, the peripheral grating suppressed the amplitude of
response fluctuations to the central spot (Fig. 6A). To quantify
the nature of this suppression, we performed an LN analysis of
the subthreshold response, similar to the analysis described
above for the spike rate. Two LN models were fit that allowed,
between conditions, a tonic membrane polarization (difference
RETINAL CIRCUITRY FOR PERIPHERAL SUPPRESSION
4335
FIG. 8. Peripheral contrast effects on subthreshold and spiking responses are correlated. A: relationship between effect of grating on average membrane
potential and reduced gain of membrane potential (n ⫽ 21 OFF cells with firing rates of at least 1 spike/s in both conditions). Change in average membrane
potential was measured as change in the y-intercept of nonlinear function (negative values indicate that peripheral grating caused hyperpolarization). Gain change
shows gain in the spot ⫹ grating condition (gs⫹g) relative to gain in the spot alone condition (gs). Correlation was not significant. B: relationship between effect
on average membrane potential and horizontal shift in spike nonlinearity (positive values indicate rightward shift in presence of grating). Correlation was
significant (P ⬍ 0.05). C: relationship between reduced gain in membrane potential and horizontal shift in spike nonlinearity. Correlation was significant (P ⬍
0.01).
J Neurophysiol • VOL
response was correlated with both the hyperpolarization (r ⫽
⫺0.57; P ⬍ 0.05) and the reduced gain in the subthreshold
response (r ⫽ ⫺0.79; P ⬍ 0.01; Fig. 8, B and C). Thus the
suppression of the spiking response was related to both suppressive effects expressed in the subthreshold potential.
Peripheral contrast increases ganglion cell membrane
conductance
We tested whether the hyperpolarization in OFF cells, caused
by the peripheral grating, could be explained by a direct
inhibitory synapse onto the ganglion cell dendrite. To test this,
we made intracellular recordings with QX-314 in the pipette to
block spiking and improve the space clamp. We measured the
response to the peripheral grating in voltage clamp at several
holding potentials. Every cell showed an increased conductance
(i.e., positive slope on the I-V plot) during both the transient and
sustained periods of the grating (Fig. 9). For OFF cells (n ⫽ 6), the
conductance increased by 8 ⫾ 2 nS for the transient response and
1.3 ⫾ 0.4 nS for the sustained response. The transient response
reversed at ⫺94 ⫾ 4 mV, and the sustained response (2–10 s after
grating onset) reversed at ⫺97 ⫾ 18 mV.
The above reversal potential measurements suggest that, for
OFF cells, the grating simultaneously increased an inhibitory
conductance (⌬gGABA/glycine) and decreased an excitatory conductance (⌬gcation), which would move the reversal potential
for the summed conductance (⌬gtotal) negative to EGABA/glycine.
To determine the relative contributions of the two underlying
conductances [where the conductances represent changes (⌬)
from resting conductances], we used the following formula
a ⫽ (E GABA/glycine – E total)/(E total – E cation)
where a is the ratio between the conductances (⌬gcation/
⌬gGABA/glycine), EGABA/glycine ⫽ ⫺68 mV, Ecation ⫽ 0 mV, and
Etotal was the measured reversal potential for the leak-subtracted response to the grating. The above equation follows
from Ohm’s law, given that, during the sustained period of the
grating, ⌬ication ⫽ ⫺⌬iGABA/glycine. From this equation, we
could divide the total conductance into the two underlying
conductances (after establishing their relative contributions).
This procedure is depicted graphically in Fig. 9C. For OFF cells
(n ⫽ 6), the grating evoked an inhibitory conductance of 1.7 ⫾
0.4 nS (P ⬍ 0.05; 2-tailed t-test) in parallel with a decreased
excitatory conductance of 0.43 ⫾ 0.14 nS (P ⬍ 0.05; Fig. 9D).
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
in the y-intercept of the nonlinearity) plus a gain change
(difference in the slope of the nonlinearity; see METHODS). The
example OFF cell shows the typical effect: the grating caused
both hyperpolarization and a reduced gain (Fig. 7A; see also
Fig. 2). The accompanying spike response is shown with the
threshold change model fitted to the data (Fig. 7A; see also Fig.
2). For ON cells, there were two patterns. In the first pattern, the
grating caused a hyperpolarization and reduced gain, similar to
the OFF cell (Fig. 7B). In the second pattern, the grating caused
a depolarization and reduced gain (Fig. 7C). The accompanying spike responses for the two ON cells are shown with the
combined model fitted to the data (Fig. 7, B and C). For both
patterns, the grating reduced the slope of the spike nonlinearity,
but the rightward shift of the spike nonlinearity was only
present in the case where the grating evoked hyperpolarization
(Fig. 7B).
Next, we compared quantitatively the reduced gain during
the spot ⫹ grating conditions across cells. For every cell
recorded intracellularly (n ⫽ 25 OFF cells and 7 ON cells), the
grating reduced the gain of the subthreshold response. For OFF
cells, the gain was reduced significantly below 1.0 to 0.68 ⫾
0.02 (P ⬍ 0.001; i.e., reduction of ⬃32%). For ON cells, the
gain was reduced to 0.78 ⫾ 0.06 (P ⬍ 0.001; i.e., reduction of
⬃22%). The change in the y-intercept is a measure of membrane polarization, independent of the nonlinearity. The yintercept showed a hyperpolarization for OFF cells (⫺1.0 ⫾ 0.2
mV; P ⬍ 0.001), similar to the direct analysis of membrane
potential above, but no significant change for ON cells (⫹0.3 ⫾ 0.4
mV). Thus the reduced gain in the membrane potential response,
observed in all cells, was not always accompanied by membrane
hyperpolarization. The apparent independence of these two inhibitory effects suggests distinct underlying cellular mechanisms.
To further analyze the relationship between the suppressive
effects on spiking and subthreshold responses, we measured
three correlations. This analysis was performed on OFF cells,
where the effect of the grating on spiking could be described
concisely as an increased threshold; the analysis was further
restricted to those OFF cells that fired at a rate of at least 1
spike/s for both conditions (n ⫽ 21 cells). There was not a
significant correlation between the reduced gain in the subthreshold response and the hyperpolarization (i.e., the change
in the y-intercept of the nonlinear function; Fig. 8A). Thus
these two effects on the subthreshold response could arise from
different cellular mechanisms; we further support this conclusion with analyses below. The rightward shift of the spike
4336
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
For ON cells (n ⫽ 4), the conductance increased by 5 ⫾ 3 nS
for the transient response and 2.2 ⫾ 0.6 nS for the sustained
response. The transient response reversed at ⫺29 ⫾ 1 mV, and
the sustained response reversed at ⫺55 ⫾ 6 mV. For the
sustained response, the grating evoked an inhibitory conductance of 1.8 ⫾ 0.5 nS (P ⬍ 0.05) in parallel with an increased
excitatory conductance of 0.41 ⫾ 0.22 nS (not significant; Fig.
9D). Therefore in all cells, the grating evoked an increased
inhibitory conductance, consistent with a direct inhibitory
synapse on the ganglion cell dendrite. This inhibitory conductance was accompanied either by a decreased excitatory conductance (OFF cells) or a trend toward an increased excitatory
conductance (ON cells) that depended on cell type.
Evidence that peripheral contrast reduces the gain of
subthreshold ganglion cell center responses by inhibiting
presynaptic bipolar cells
We next tested whether the reduced gain of the subthreshold
response could be explained by postsynaptic shunting inhibition or rather by presynaptic inhibition of bipolar terminals. To
test this, we made intracellular recordings with QX-314 in the
pipette, to block spiking and improve space clamp, and compared recordings of membrane current (Im; with the holding
potential, Vhold, near the resting potential) to recordings of
membrane voltage (Vm). Consider first the voltage-clamp condition, where conductances in parallel would add. We considered the total membrane conductance as a sum of three conductances: synaptic conductance driven by the central spot
(gcenter), synaptic conductance driven by the peripheral grating
(ggrating), and a leak term (gleak), each with an associated
reversal potential
I m ⫽ g center(V hold – E center) ⫹ g grating(V hold – E grating) ⫹ g leak(V hold – E leak)
If the peripheral grating acts purely by postsynaptic inhibition,
its effect would be exclusively on ggrating. During the analysis
J Neurophysiol • VOL
period (2–10 s after grating onset or offset), we expect ggrating
and gleak to be relatively steady (Fig. 9), and we assume all
reversal potentials to be constant. Thus an effect on ggrating
would evoke a tonic offset in current between the two conditions, with no additional effect on the gain of the center
response, because the modulation amplitude of gcenter would
remain the same. If instead the peripheral grating acts through
inhibition of the presynaptic bipolar terminal, in the equation
above, the modulation amplitude of gcenter itself would change,
which should evoke a reduced gain under voltage clamp
similar to that observed under current clamp. More generally,
postsynaptic shunting inhibition, caused by tonic inhibitory
synapses, should only be present under current clamp.
We used the LN model above to compare the effects of the
grating on Im and Vm (Fig. 10). Across 13 cells (n ⫽ 9 OFF cells
and 4 ON cells), the reduction in gain was the same in both
conditions: the grating reduced the center gain to 0.73 ⫾ 0.04
(i.e., ⬃27% reduction) for Im and to 0.72 ⫾ 0.04 (i.e., 28%
reduction) for Vm (Fig. 10). The effect was similar for OFF cells
(Vm, 0.73 ⫾ 0.03; Im 0.74 ⫾ 0.03) and ON cells (Vm, 0.70 ⫾
0.10; Im 0.69 ⫾ 0.10). These gain reductions in Vm are
consistent with those recorded with standard pipette solution
above (⬃22–32%). The reduced gain was accompanied in OFF
cells by a slight hyperpolarization as reflected by a change in
the y-intercept of the nonlinearity of ⫺0.6 ⫾ 0.5 mV (n ⫽ 9).
This hyperpolarization was somewhat small and inconsistent
across cells, compared with recordings above, perhaps caused by
the more depolarized value for EGABA/glycine (i.e., with Br⫺ in the
pipette solution) and the accompanying decreased driving force
on inhibition (see METHODS). Under voltage clamp, OFF cells
showed a small outward current (16 ⫾ 13 pA). For ON cells, there
was little change in the y-intercept under either condition (current
clamp: ⫺0.2 ⫾ 0.7 mV; voltage clamp: 7 ⫾ 28 pA). In general,
our main conclusion from the comparison of current-clamp and
voltage-clamp recordings is based on the similar reduction in gain
across the two conditions. The reduced gain of the ganglion cell
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
FIG. 9. Peripheral contrast causes an increased membrane conductance. A1: OFF cell response to peripheral grating was measured in voltage clamp: Vhold ⫽
⫺58 or ⫺74 mV. Traces are leak-subtracted and show response to a single presentation of grating. A2: I-V plots show amplitude of transient current, right after
grating onset (dark symbol), and sustained current during main period of analysis, 2–10 s after grating onset (white symbols). Lines show regression fits;
x-intercept of fit is apparent reversal potential. This reversal was ⫺95 (transient response) or ⫺90 mV (sustained response). B1: same format as A1 for an ON
cell: Vhold ⫽ ⫺37 or ⫺63 mV. B2: same format as A2 for ON cell in B1. Reversal was ⫺32 (transient response) or ⫺71 mV (sustained response). C: sustained
response from A2 is shown with calculated underlying excitatory (dashed gray line) and inhibitory conductances (solid gray line). The 2 underlying conductances,
when summed, equal total conductance (i.e., regression fit to data; solid black line). D: average excitatory conductance (gex) and inhibitory conductance (gin)
during sustained response to grating for OFF cells (n ⫽ 6) and ON cells (n ⫽ 4). Error bars indicate ⫾SE.
RETINAL CIRCUITRY FOR PERIPHERAL SUPPRESSION
4337
center Vm response must be primarily caused by amacrine cell
inhibition of presynaptic bipolar cell terminals.
DISCUSSION
We measured the effect of peripheral suppression on the
center response of mammalian retinal ganglion cells using
intracellular recording. We used a peripheral grating with a
spatial frequency that was beyond the resolution of the horizontal cell network to selectively stimulate amacrine cells (Fig.
3). Our main finding was that the peripheral grating caused two
suppressive effects on the ganglion cell subthreshold membrane potential: a reduced gain of the center response, evident
in all cells (Figs. 7 and 8), and a tonic membrane hyperpolarization, most consistently observed in OFF cells (Figs. 6 – 8).
The tonic membrane polarization was consistent with an increased inhibitory conductance at the ganglion cell dendrite in
parallel with a decreased (OFF cells) or increased (ON cells)
excitatory conductance (Fig. 9). For both cell types, the reduced gain of the center response was similar under currentclamp and voltage-clamp conditions, which suggests that this
reduced gain reflects an inhibition of presynaptic bipolar terminals (Fig. 10).
Circuitry of amacrine-mediated peripheral suppression
A model explaining the main results is shown in Fig. 11. The
drifting grating in the periphery stimulates bipolar cells, which
each have a nonlinearity at their synaptic output (Demb et al.
2001a; Enroth-Cugell and Freeman 1987). Each bipolar cell
J Neurophysiol • VOL
then acts as a nonlinear subunit, which can increase its release
more than it can decrease its release (i.e., rectification). The
drifting bars stimulate the subunits asynchronously, which
leads to a steady increase in the summed excitatory drive onto
the amacrine cell (Demb et al. 2001a; Enroth-Cugell and
Robson 1966; Hochstein and Shapley 1976; Olveczky et al.
2003). Tonic stimulation of the amacrine cell drives tonic
inhibition of the recorded ganglion cell and its presynaptic
bipolar cells. The long-range inhibitory signal is carried by a
spiking amacrine cell, which fires conventional sodium spikes,
FIG. 11. Circuit model to explain influence of long-range amacrine signaling on ganglion cell center response. For OFF cells, amacrine cell inhibition acts
at 2 points: a synapse onto the ganglion cell dendrite and a synapse onto the
presynaptic bipolar cell terminal. Both synapses result in tonic hyperpolarization of the ganglion cell. Synapse on the bipolar terminal further causes a
reduced gain of center response to the spot. For ON cells, the circuit would be
similar except that, under these conditions, grating sometimes caused a tonic
depolarization that could be explained by inadvertent stimulation of bipolar
cells at the edge of the ganglion cell dendritic tree (arrow pointing from bipolar
terminal to extended ganglion cell dendrite, in dashed lines).
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
FIG. 10. Peripheral contrast reduces gain
of subthreshold response similarly in membrane voltages and currents. A: time-course of
effect of peripheral grating for an OFF cell
measured in current clamp or voltage clamp
(Vhold ⫽ ⫺73 mV); same format as Fig. 6A.
Recording solution included QX-314 to block
sodium channels. B: linear-nonlinear model
for OFF cell in A, measured in current clamp or
voltage clamp. Grating reduced gain to a similar degree under current clamp (49%) and
voltage clamp (46%). Fits to nonlinear functions are based on depolarizing responses (current clamp) or inward currents (voltage
clamp). Note that for voltage-clamp recording
here and in C, the y-axis of linear filter and
both axes of nonlinearity have been plotted in
reverse (i.e., negative values are going upward
or rightward) to facilitate comparison with LN
model for current-clamp recording. C: same as
B for an ON cell. Grating reduced gain to a
similar degree under current clamp (13%) and
voltage clamp (18%; Vhold ⫽ ⫺47 mV). D:
across cells (n ⫽ 9 OFF cells, n ⫽ 4 ON cells),
reduced gains measured under voltage-clamp
and current-clamp conditions were similar.
Gain change was measured in the spot ⫹
grating condition (gs⫹g) relative to gain in the
spot alone condition (gs). Points lie near identity line.
4338
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
Werblin and Copenhagen 1974). In mudpuppy, a spinning
windmill stimulus, like the drifting grating stimulus in this
study, was shown to stimulate amacrine cells but not horizontal
cells. The windmill stimulus, presented in the receptive field
periphery, suppressed the spiking response of ganglion cells by
causing a membrane hyperpolarization. However, in several
studies, the windmill apparently did not suppress bipolar cells
(Thibos and Werblin 1978; Werblin 1972; Werblin and Copenhagen 1974). Thus in mudpuppy, the amacrine cells seem to act
mostly at the ganglion cell dendrite, whereas we showed here
in a mammal an important role for suppression of the presynaptic bipolar terminals.
The amacrine cells that mediate the windmill-evoked suppression rely on sodium action potentials to convey signals
laterally over long distances (Cook and Werblin 1994; Cook et
al. 1998). Bath-applied TTX blocked suppression in ganglion
cells but not bipolar cells in one study (Cook and McReynolds
1998; Cook et al. 1998). However, another study showed that
inhibition at salamander bipolar terminals is TTX-sensitive
(Shields and Lukasiewicz 2003). Thus there may be a role for
presynaptic inhibition in the salamander circuit for peripheral
suppression similar to the presynaptic mechanism shown here.
In ON-OFF ganglion cells, the OFF pathway can be selectively
suppressed during transient shifts of a grating in the surround,
consistent with amacrine-mediated suppression of presynaptic
OFF bipolar terminals (Geffen et al. 2007).
Asymmetry between
ON
and
OFF
pathways
Asymmetries exist between the parallel ON and OFF ganglion
cell pathways, for example, in receptive field size and the
pattern of excitatory and inhibitory synaptic input (Chichilnisky and Kalmar 2002; Murphy and Rieke 2006; Pang et al.
2003; Sagdullaev et al. 2006; Zaghloul et al. 2003). Here, we
found that, in the presence of the grating, OFF cells showed a
more consistent suppression of their spike rates and a more
consistent membrane hyperpolarization. However, these asymmetries must be interpreted with some caution. For example,
we did not systematically explore multiple spatial and temporal
frequency conditions for the surround grating, and doing so
might reveal conditions that evoke larger suppressive effects in
ON cells. Indeed, there were clear cases where ON cells showed
suppressive effects, including suppressed spike rates, reduced
gain of the subthreshold response and membrane hyperpolarizations (Figs. 4 –7). We conclude that different amacrine cell
pathways interact with ON and OFF Y cells (i.e., there is not a
single ON-OFF amacrine pathway that inhibits both ganglion cell
circuits). Consistent with this, in the in vivo rabbit retina,
picrotoxin (GABAA/C receptor antagonist) blocked the suppressive effect of a peripheral windmill stimulus for OFF-center
Y cells but not ON-center Y cells, which also predicts the
involvement of two distinct amacrine cell mechanisms (Caldwell and Daw 1978a).
Peripheral suppression throughout the visual system
Comparing the circuit for peripheral suppression between
mammals and lower vertebrates
The circuit revealed here in a mammalian retina shows both
similarities and differences to a circuit proposed in lower
vertebrate retina (Thibos and Werblin 1978; Werblin 1972;
J Neurophysiol • VOL
Mechanisms for peripheral suppression appear first in the
retina but are repeated at several stages of the visual pathway.
For example, cells in the lateral geniculate nucleus (LGN) of
the thalamus show peripheral suppression (Bonin et al. 2005;
Solomon et al. 2002; Webb et al. 2005b). Some of the sup-
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
blocked by TTX (Demb et al. 1999, 2001a; Flores-Herr et al.
2001; Olveczky et al. 2003; Roska and Werblin 2003; Taylor
1999). These amacrine cells are presumably those that extend
axons millimeters across the retina (Dacey 1989; Famiglietti
1992; Stafford and Dacey 1997; Vaney et al. 1988). These
long-range amacrine cells represent a class of cell that includes
several types (Lin and Masland 2006; Volgyi et al. 2001). The
specific types that synapse onto Y-type cells and their presynaptic bipolar terminals are unknown.
The reduced gain of the center response in the subthreshold
membrane potential is apparently driven by amacrine cell
inhibition of the central bipolar terminals (i.e., those conveying
the spot response; Fig. 11). The hyperpolarization, shown most
consistently in OFF cells, is driven by a combination of direct
inhibition of the ganglion cell and inhibition of tonic glutamate
release from presynaptic bipolar terminals (Figs. 9 and 11). For
ON cells, the grating sometimes caused a slight depolarization
that could be explained if the edge of the grating sometimes
inadvertently caused direct stimulation of bipolar cells at the
edge of the receptive field center (Figs. 6, 7, and 11). This may
have occurred more frequently in ON cells because they are
relatively larger than OFF cells in guinea pig retina, as found in
primate and human parasol cells (Chichilnisky and Kalmar
2002; Dacey and Petersen 1992; J. Demb, unpublished observations). This inadvertent center stimulation may have also
occurred in a few OFF cells and explain why the grating did not
cause a tonic membrane hyperpolarization in all cases (Fig.
8B). Despite this variability in tonic membrane polarization,
the grating reduced the gain of the center response in the
subthreshold membrane potential for every ON and OFF cell
studied. This suggests that the peripheral grating always
evoked an inhibitory effect on those central-most bipolar terminals that conveyed the spot response.
Our data support two sites of synaptic inhibition driven by
long-range amacrine cells: bipolar terminals and ganglion
cells. This conclusion is consistent with previous studies in
mammalian retina (Flores-Herr et al. 2001; Roska and Werblin
2003; Taylor 1999). However, we added to these previous
efforts in four regards. First, we used a high spatial frequency
surround stimulus and showed that it does not stimulate horizontal cells (Fig. 3). Thus surround effects studied here can be
ascribed specifically to amacrine cells. Second, we used the LN
model to study effects of amacrine cell stimulation on the
temporal tuning of the center response. Under our conditions,
effects on the temporal tuning were minimal. The grating
caused a slightly more biphasic filter in OFF cells, which should
attenuate low temporal frequencies (Fig. 5). Third, we used the
LN model to quantify gain changes separately from static
nonlinear (i.e., rectifying) influences on membrane voltage or
current. Fourth, by directly comparing voltage-clamp and current-clamp recordings in the same cell, we determined whether
shunting of the ganglion cell leads to reduced gain of the center
response. The shunting effect was shown to be negligible (Fig.
10).
RETINAL CIRCUITRY FOR PERIPHERAL SUPPRESSION
pression at the LGN can presumably be explained by a retinal
mechanism, but other central mechanisms may also play a role.
Peripheral suppression also exists in the primary visual cortex,
where its orientation sensitivity, binocular nature and timecourse suggest an intracortical mechanism (Bair et al. 2003;
DeAngelis et al. 1994; Smith et al. 2006; Webb et al. 2005a).
Further surround effects appear in extrastriate areas (see Albright and Stoner 2002; Allman et al. 1985). At all stages, the
apparent role of peripheral suppression is to create a context in
which to interpret the strength of a central stimulus. It will be
interesting to learn whether the circuitry for peripheral suppression described here is repeated at other stages.
ACKNOWLEDGMENTS
We thank D. Green and V. Bonin for comments on the manuscript.
Present address of K. Boahen: Department of Bioengineering, Stanford
University, Palo Alto, CA.
GRANTS
REFERENCES
Albright TD, Stoner GR. Contextual influences on visual processing. Annu
Rev Neurosci 25: 339 –379, 2002.
Allman J, Miezin F, McGuinness E. Stimulus specific responses from
beyond the classical receptive field: neurophysiological mechanisms for
local-global comparisons in visual neurons. Annu Rev Neurosci 8: 407– 430,
1985.
Baccus SA, Meister M. Fast and slow contrast adaptation in retinal circuitry.
Neuron 36: 909 –919, 2002.
Bair W, Cavanaugh JR, Movshon JA. Time course and time-distance
relationships for surround suppression in macaque V1 neurons. J Neurosci
23: 7690 –7701, 2003.
Baldridge WH, Vaney DI, Weiler R. The modulation of intercellular coupling in the retina. Semin Cell Dev Biol 9: 311–318, 1998.
Beaudoin DL, Borghuis BG, Demb JB. Cellular basis of contrast gain control
over the receptive field center of mammalian retinal ganglion cells. J Neurosci 27: 2636 –2645, 2007.
Bonin V, Mante V, Carandini M. The suppressive field of neurons in lateral
geniculate nucleus. J Neurosci 25: 10844 –10856, 2005.
Bormann J, Hamill OP, Sakmann B. Mechanism of anion permeation
through channels gated by glycine and gamma-aminobutyric acid in mouse
cultured spinal neurones. J Physiol 385: 243–286, 1987.
Brainard DH. The psychophysics toolbox. Spat Vis 10: 433– 436, 1997.
Caldwell JH, Daw NW. Effects of picrotoxin and strychnine on rabbit retinal
ganglion cells: changes in centre surround receptive fields. J Physiol 276:
299 –310, 1978a.
Caldwell JH, Daw NW. New properties of rabbit retinal ganglion cells.
J Physiol 276: 257–276, 1978b.
Carandini M, Demb JB, Mante V, Tolhurst DJ, Dan Y, Olshausen BA,
Gallant JL, Rust NC. Do we know what the early visual system does?
J Neurosci 25: 10577–10597, 2005.
Chander D, Chichilnisky EJ. Adaptation to temporal contrast in primate and
salamander retina. J Neurosci 21: 9904 –9916, 2001.
Chichilnisky EJ. A simple white noise analysis of neuronal light responses.
Network 12: 199 –213, 2001.
Chichilnisky EJ, Kalmar RS. Functional asymmetries in ON and OFF ganglion
cells of primate retina. J Neurosci 22: 2737–2747, 2002.
Cleland BG, Levick WR. Brisk and sluggish concentrically organized ganglion cells in the cat’s retina. J Physiol 240: 421– 456, 1974.
Cohen ED. Synaptic mechanisms shaping the light-response in retinal ganglion cells. Prog Brain Res 131: 215–228, 2001.
Cook PB, Lukasiewicz PD, McReynolds JS. Action potentials are required
for the lateral transmission of glycinergic transient inhibition in the amphibian retina. J Neurosci 18: 2301–2308, 1998.
Cook PB, McReynolds JS. Lateral inhibition in the inner retina is important
for spatial tuning of ganglion cells. Nat Neurosci 1: 714 –719, 1998.
J Neurophysiol • VOL
Cook PB, Werblin FS. Spike initiation and propagation in wide field transient
amacrine cells of the salamander retina. J Neurosci 14: 3852–3861, 1994.
Dacey DM. Axon-bearing amacrine cells of the macaque monkey retina.
J Comp Neurol 284: 275–293, 1989.
Dacey D, Packer OS, Diller L, Brainard D, Peterson B, Lee B. Center
surround receptive field structure of cone bipolar cells in primate retina.
Vision Res 40: 1801–1811, 2000.
Dacey DM, Petersen MR. Dendritic field size and morphology of midget and
parasol ganglion cells of the human retina. Proc Natl Acad Sci USA 89:
9666 –9670, 1992.
DeAngelis GC, Freeman RD, Ohzawa I. Length and width tuning of neurons
in the cat’s primary visual cortex. J Neurophysiol 71: 347–374, 1994.
Demb JB, Haarsma L, Freed MA, Sterling P. Functional circuitry of the
retinal ganglion cell’s nonlinear receptive field. J Neurosci 19: 9756 –9767,
1999.
Demb JB, Zaghloul K, Haarsma L, Sterling P. Bipolar cells contribute to
nonlinear spatial summation in the brisk-transient (Y) ganglion cell in
mammalian retina. J Neurosci 21: 7447–7454, 2001a.
Demb JB, Zaghloul K, Sterling P. Cellular basis for the response to
second-order motion cues in y retinal ganglion cells. Neuron 32: 711–721,
2001b.
Dhingra NK, Kao Y-H, Sterling P, Smith RG. Contrast threshold of a
brisk-transient ganglion cell in vitro. J Neurophysiol 89: 2360 –2369, 2003.
Duebel J, Haverkamp S, Schleich W, Feng G, Augustine GJ, Kuner T,
Euler T. Two-photon imaging reveals somatodendritic chloride gradient in
retinal ON-type bipolar cells expressing the biosensor Clomeleon. Neuron
49: 81–94, 2006.
Enroth-Cugell C, Freeman AW. The receptive-field spatial structure of cat
retinal Y cells. J Physiol 384: 49 –79, 1987.
Enroth-Cugell C, Jakiela HG. Suppression of cat retinal ganglion cell
responses by moving patterns. J Physiol 302: 49 –72, 1980.
Enroth-Cugell C, Robson JG. The contrast sensitivity of retinal ganglion
cells of the cat. J Physiol 187: 517–552, 1966.
Famiglietti EV. Polyaxonal amacrine cells of rabbit retina: morphology and
stratification of PA1 cells. J Comp Neurol 316: 391– 405, 1992.
Flores-Herr N, Protti DA, Wassle H. Synaptic currents generating the
inhibitory surround of ganglion cells in the mammalian retina. J Neurosci
21: 4852– 4863, 2001.
Frishman LJ, Linsenmeier RA. Effects of picrotoxin and strychnine on
non-linear responses of Y-type cat retinal ganglion cells. J Physiol 324:
347–363, 1982.
Geffen MN, de Vries SEJ, Meister M. Retinal ganglion cells can rapidly
change polarity from off to on. PLoS Biol 5: e65, 2007.
Hochstein S, Shapley RM. Linear and nonlinear spatial subunits in Y cat
retinal ganglion cells. J Physiol 262: 265–284, 1976.
Hu EH, Dacheux RF, Bloomfield SA. A flattened retina-eyecup preparation
suitable for electrophysiological studies of neurons visualized with transscleral infrared illumination. J Neurosci Methods 103: 209 –216, 2000.
Kier CK, Buchsbaum G, Sterling P. How retinal microcircuits scale for
ganglion cells of different size. J Neurosci 15: 7673–7683, 1995.
Kim KJ, Rieke F. Temporal contrast adaptation in the input and output signals
of salamander retinal ganglion cells. J Neurosci 21: 287–299, 2001.
Kuffler SW. Discharge patterns and functional organization of mammalian
retina. J Neurophysiol 16: 37– 68, 1953.
Lankheet MJ, Prickaerts JH, van de Grind WA. Responses of cat horizontal cells to sinusoidal gratings. Vision Res 32: 997–1008, 1992.
Lee YW, Schetzen M. Measurement of the Wiener kernels of a nonlinear
system by cross-correlation. Int J Control 2: 237–254, 1965.
Lin B, Masland RH. Populations of wide-field amacrine cells in the mouse
retina. J Comp Neurol 499: 797– 809, 2006.
Mangel SC. Analysis of the horizontal cell contribution to the receptive field
surround of ganglion cells in the rabbit retina. J Physiol 442: 211–234, 1991.
Manookin M, Demb JB. Presynaptic mechanism for slow contrast adaptation
in mammalian retinal ganglion cells. Neuron 50: 453– 464, 2006.
Marmarelis PZ, Marmarelis VZ. Analysis of Physiological Systems: The
White-Noise Approach. New York: Plenum Press, 1978.
Masland RH. The fundamental plan of the retina. Nat Neurosci 4: 877– 886,
2001.
McMahon MJ, Packer OS, Dacey DM. The classical receptive field surround
of primate parasol ganglion cells is mediated primarily by a non-GABAergic
pathway. J Neurosci 24: 3736 –3745, 2004.
Murphy GJ, Rieke F. Network variability limits stimulus-evoked spike
timing precision in retinal ganglion cells. Neuron 52: 511–524, 2006.
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
This study was supported by National Eye Institute Grants EY-07003,
T32-EY-13934, EY-07035, EY-00828, and EY-14454 and the Packard Foundation. J. B. Demb is supported by a Research to Prevent Blindness Career
Development Award and an Alfred P. Sloan Research Fellowship.
4339
4340
ZAGHLOUL, MANOOKIN, BORGHUIS, BOAHEN, AND DEMB
J Neurophysiol • VOL
Solomon SG, White AJ, Martin PR. Extraclassical receptive field properties
of parvocellular, magnocellular, and koniocellular cells in the primate lateral
geniculate nucleus. J Neurosci 22: 338 –349, 2002.
Stafford DK, Dacey D. Physiology of the A1 amacrine: a spiking, axonbearing interneuron of the macaque monkey retina. Vis Neurosci 14:
507–522, 1997.
Sterling P, Demb JB. Retina. In: Synaptic Organization of the Brain (5th ed.),
edited by Shephard G. New York: Oxford, 2004, p. 217–269.
Taylor WR. TTX attenuates surround inhibition in rabbit retinal ganglion
cells. Vis Neurosci 16: 285–290, 1999.
Thibos LN, Werblin FS. The properties of surround antagonism elicited by
spinning windmill patterns in the mudpuppy retina. J Physiol 278: 101–116,
1978.
Vaney DI, Peichl L, Boycott BB. Neurofibrillar long-range amacrine cells in
mammalian retinae. Proc R Soc Lond B Biol Sci 235: 203–219, 1988.
Victor JD. The dynamics of the cat retinal X cell centre. J Physiol 386:
219 –246, 1987.
Volgyi B, Xin D, Amarillo Y, Bloomfield SA. Morphology and physiology of
the polyaxonal amacrine cells in the rabbit retina. J Comp Neurol 440:
109 –125, 2001.
Wässle H. Parallel processing in the mammalian retina. Nat Rev Neurosci 5:
747–757, 2004.
Webb BS, Dhruv NT, Solomon SG, Tailby C, Lennie P. Early and late
mechanisms of surround suppression in striate cortex of macaque. J Neurosci 25: 11666 –11675, 2005a.
Webb BS, Tinsley CJ, Vincent CJ, Derrington AM. Spatial distribution of
suppressive signals outside the classical receptive field in lateral geniculate
nucleus. J Neurophysiol 94: 1789 –1797, 2005b.
Werblin FS. Lateral interactions at inner plexiform layer of vertebrate retina:
antagonistic responses to change. Science 175: 1008 –1010, 1972.
Werblin FS, Copenhagen DR. Control of retinal sensitivity. 3. Lateral
interactions at the inner plexiform layer. J Gen Physiol 63: 88 –110,
1974.
Wiener N.Nonlinear Problems in Random Theory. New York: Wiley, 1958.
Yin L, Smith RG, Sterling P, Brainard DH. Chromatic properties of
horizontal and ganglion cell responses follow a dual gradient in cone opsin
expression. J Neurosci 26: 12351–12361, 2006.
Zaghloul KA, Boahen K, Demb JB. Different circuits for ON and OFF retinal
ganglion cells cause different contrast sensitivities. J Neurosci 23: 2645–
2654, 2003.
Zaghloul KA, Boahen K, Demb JB. Contrast adaptation in subthreshold and
spiking responses of mammalian Y-type retinal ganglion cells. J Neurosci
25: 860 – 868, 2005.
97 • JUNE 2007 •
www.jn.org
Downloaded from jn.physiology.org on June 12, 2007
O’Brien BJ, Isayama T, Richardson R, Berson DM. Intrinsic physiological
properties of cat retinal ganglion cells. J Physiol 538: 787– 802, 2002.
Olveczky BP, Baccus SA, Meister M. Segregation of object and background
motion in the retina. Nature 423: 401– 408, 2003.
Pang JJ, Gao F, Wu SM. Light-evoked excitatory and inhibitory synaptic
inputs to ON and OFF alpha ganglion cells in the mouse retina. J Neurosci 23:
6063– 6073, 2003.
Passaglia CL, Enroth-Cugell C, Troy JB. Effects of remote stimulation on the
mean firing rate of cat retinal ganglion cells. J Neurosci 21: 5794 –5803, 2001.
Peichl L, Gonzalez-Soriano J. Unexpected presence of neurofilaments in
axon-bearing horizontal cells of the mammalian retina. J Neurosci 13:
4091– 4100, 1993.
Peichl L, Ott H, Boycott BB. Alpha ganglion cells in mammalian retinae.
Proc R Soc Lond B Biol Sci 231: 169 –197, 1987.
Pelli DG. The VideoToolbox software for visual psychophysics, transforming
numbers into movies. Spat Vis 10: 437– 442, 1997.
Rieke F. Temporal contrast adaptation in salamander bipolar cells. J Neurosci
21: 9445–9454, 2001.
Robertson B. Characteristics of GABA-activated chloride channels in mammalian dorsal root ganglion neurones. J Physiol 411: 285–300, 1989.
Rodieck RW. Quantitative analysis of cat retinal ganglion cell response to
visual stimuli. Vision Res 5: 583– 601, 1965.
Roska B, Werblin F. Vertical interactions across ten parallel, stacked representations in the mammalian retina. Nature 410: 583–587, 2001.
Roska B, Werblin F. Rapid global shifts in natural scenes block spiking in
specific ganglion cell types. Nat Neurosci 6: 600 – 608, 2003.
Sagdullaev BT, McCall MA, Lukasiewicz PD. Presynaptic inhibition modulates spillover, creating distinct dynamic response ranges of sensory output.
Neuron 50: 923–935, 2006.
Sakai HM, Naka K. Signal transmission in the catfish retina. IV. Transmission to ganglion cells. J Neurophysiol 58: 1307–1328, 1987.
Sakai HM, Wang JL, Naka K. Contrast gain control in the lower vertebrate
retinas. J Gen Physiol 105: 815– 835, 1995.
Schwartz EA. Organization of on-off cells in the retina of the turtle. J Physiol
230: 1–14, 1973.
Shapley RM, Victor JD. Nonlinear spatial summation and the contrast gain
control of cat retinal ganglion cells. J Physiol 290: 141–161, 1979.
Shields CR, Lukasiewicz PD. Spike-dependent GABA inputs to bipolar cell
axon terminals contribute to lateral inhibition of retinal ganglion cells.
J Neurophysiol 89: 2449 –2458, 2003.
Smith MA, Bair W, Movshon JA. Dynamics of suppression in macaque
primary visual cortex. J Neurosci 26: 4826 – 4834, 2006.
Solomon SG, Lee BB, Sun H. Suppressive surrounds and contrast gain in
magnocellular-pathway retinal ganglion cells of macaque. J Neurosci 26:
8715– 8726, 2006.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement